f16 and pm4f are always the same swizzle, so remove adaptor

src/core/SkXfermodeU64.cpp

Index: src/core/SkXfermodeU64.cpp
diff --git a/src/core/SkXfermodeU64.cpp b/src/core/SkXfermodeU64.cpp
index 630a9d068351f103f22a8d9ccb3770bd18a125eb..2cf3e1b97ed370372ea3c3015f5bb6fd19101abb 100644
--- a/src/core/SkXfermodeU64.cpp
+++ b/src/core/SkXfermodeU64.cpp
@@ -50,15 +50,6 @@ template <DstType D> uint64_t store_to_dst(const Sk4f& x4) {
     return (D == kU16_Dst) ? store_to_u16(x4) : SkFloatToHalf_01(x4);
 }
 
-static inline Sk4f pm_to_rgba_order(const Sk4f& x) {
-    if (SkPM4f::R == 0) {
-        return x;   // we're already RGBA
-    } else {
-        // we're BGRA, so swap R and B
-        return SkNx_shuffle<2, 1, 0, 3>(x);
-    }
-}
-
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
 template <DstType D> void xfer_u64_1(const SkXfermode* xfer, uint64_t dst[],
@@ -135,7 +126,7 @@ const SkXfermode::D64Proc gProcs_Clear[] = {
 
 template <DstType D> void src_1(const SkXfermode*, uint64_t dst[],
                                 const SkPM4f* src, int count, const SkAlpha aa[]) {
-    const Sk4f s4 = pm_to_rgba_order(unit_to_bias<D>(Sk4f::Load(src->fVec)));
+    const Sk4f s4 = unit_to_bias<D>(Sk4f::Load(src->fVec));
     if (aa) {
         for (int i = 0; i < count; ++i) {
             const Sk4f d4 = load_from_dst<D>(dst[i]);
@@ -150,13 +141,13 @@ template <DstType D> void src_n(const SkXfermode*, uint64_t dst[],
                                 const SkPM4f src[], int count, const SkAlpha aa[]) {
     if (aa) {
         for (int i = 0; i < count; ++i) {
-            const Sk4f s4 = pm_to_rgba_order(unit_to_bias<D>(Sk4f::Load(src[i].fVec)));
+            const Sk4f s4 = unit_to_bias<D>(Sk4f::Load(src[i].fVec));
             const Sk4f d4 = load_from_dst<D>(dst[i]);
             dst[i] = store_to_dst<D>(lerp_by_coverage(s4, d4, aa[i]));
         }
     } else {
         for (int i = 0; i < count; ++i) {
-            const Sk4f s4 = pm_to_rgba_order(unit_to_bias<D>(Sk4f::Load(src[i].fVec)));
+            const Sk4f s4 = unit_to_bias<D>(Sk4f::Load(src[i].fVec));
             dst[i] = store_to_dst<D>(s4);
         }
     }
@@ -181,7 +172,7 @@ const SkXfermode::D64Proc gProcs_Dst[] = {
 
 template <DstType D> void srcover_1(const SkXfermode*, uint64_t dst[],
                                     const SkPM4f* src, int count, const SkAlpha aa[]) {
-    const Sk4f s4 = pm_to_rgba_order(Sk4f::Load(src->fVec));
+    const Sk4f s4 = Sk4f::Load(src->fVec);
     const Sk4f dst_scale = Sk4f(1 - get_alpha(s4));
     const Sk4f s4bias = unit_to_bias<D>(s4);
     for (int i = 0; i < count; ++i) {
@@ -198,7 +189,7 @@ template <DstType D> void srcover_1(const SkXfermode*, uint64_t dst[],
 template <DstType D> void srcover_n(const SkXfermode*, uint64_t dst[],
                                     const SkPM4f src[], int count, const SkAlpha aa[]) {
     for (int i = 0; i < count; ++i) {
-        const Sk4f s4 = pm_to_rgba_order(Sk4f::Load(src[i].fVec));
+        const Sk4f s4 = Sk4f::Load(src[i].fVec);
         const Sk4f dst_scale = Sk4f(1 - get_alpha(s4));
         const Sk4f s4bias = unit_to_bias<D>(s4);
         const Sk4f d4bias = load_from_dst<D>(dst[i]);